KMS Chongqing Institute of Green and Intelligent Technology, CAS
| Correlation filter tracking with complementary features | |
Wang, Wei1,2; Li, Weiguang1,2; Shi, Mingquan1
| |
| 2018 | |
| 摘要 | Although Correlation Filters (CF) tracking algorithms have inherent capability to tackle various challenging scenarios individually, none of them are robust enough to handle all the challenges simultaneously. For any online tracking based on Correlation Filters, feature is one of the most important factors due to its representation power of target appearance. In this paper, we proposed a new tracking framework by integrating the advantage of complementary features to achieve robust tracking performance. The key issue of this work lies in the fact that different features respond to different tracking challenges, which also applies to deep learning features and hand-craft features. Moreover, for the tracking speed balance, we train a light-weight deep CNN features by using end-to-end learning method, which has the same Parameter magnitude as the hand-crafted features. Experimental results on OTB-2013, OTB-2015 large benchmarks datasets show that the proposed tracker performs favorably against several state-of-the-art methods. © Springer Nature Switzerland AG 2018. |
| 语种 | 英语 |
| DOI | 10.1007/978-3-030-04224-0_42 |
| 会议(录)名称 | 25th International Conference on Neural Information Processing, ICONIP 2018
 |
| 页码 | 488-500 |
| 收录类别 | EI |
| 会议地点 | Siem Reap, Cambodia |
| 会议日期 | December 13, 2018 - December 16, 2018 |
